A new study from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) demonstrates a significant advancement in AI-powered image generation. The research focuses on improving the ability of text-to-image models, like Stable Diffusion, to follow detailed, multi-object spatial instructions. Current models often struggle with accurately placing multiple objects in specified locations and relationships within a scene. The MIT team’s new method, called ‘LayoutGPT,’ uses large language models to interpret scene descriptions and generate a detailed layout in a text-based format. This layout is then used to guide the image generation process, resulting in images that more faithfully adhere to the user’s complex spatial requests. The approach shows marked improvements in accurately depicting scenes with multiple objects in precise arrangements, a key step toward more controllable and reliable AI image synthesis. For the full details, read the article at https://technologyreview.com/2024/07/18/1094907/mit-ai-image-generation-spatial-reasoning-layoutgpt/